cross-provider 开源项目推荐（17 个）

BinaryHB0916 /iSparto

面向单人创业者的开源跨提供商 AI 开发团队，基于 Claude Code + Codex CLI 构建。Claude 规划并审查 GPT 构建的内容；一次零上下文 GPT 处理会对该计划进行盲审。一条命令，整支军队。

StrengW /cc-switch-codex-cross-provider-bridge

让 Codex 在 CC Switch 的 OpenAI Official 与第三方 Provider 之间切换后，仍能继续原来的会话。Continue Codex sessions across OpenAI Official and third-party providers behind CC Switch.

tznthou /claude-prism

Cross-provider AI code review for Claude Code — evidence-based confidence scoring with Codex, Gemini & Claude

Florious95 /team-agent

Talk once, ship a team. A multi-agent runtime for Claude Code and Codex CLI where the lead does the orchestration — across providers, in conversation.

WDahah /portable-dsh-multi-agent-plugin

One model writes the work, a different one reviews it, and neither decides whether it passed. Cross-provider AI review with verifiable evidence, for DeepSeek Harness/Cordis hosts. 中文文档见 README.zh.md。

0xDarkMatter /fleetflow

Heterogeneous cross-provider agent fleets from one orchestrator session — GLM, Codex, Grok, Pi, and Anthropic models as OS-process workers with adversarial cross-model verification, journalled resume, and a machine-wide dashboard.

Aetheria-LabsJP /puppetllm

Fake Anthropic / Bedrock (InvokeModel, Converse, batch inference, bundled S3) / OpenAI server for zero-cost human/AI-in-the-loop LLM debugging: scripted test harness (scenario rules, latency / rate-limit simulation, pytest fixture), cross-provider relay, observability (history API, JSON logs, Prometheus metrics, cache diagnostics, cost provenance)

shawnjason /Hamiltonian-Microscope

Machine-checked Lean 4 proofs for "Minimal Topologies of Forward-Local Failure in AI Systems: The Hamiltonian Microscope" — hamiltonian-path as abstract ConstraintSystem instance (inheriting Sudoku-Microscope theorems) and Layer 2 silent-commit signature formalization with provider-agnostic well-definedness.

Pher217 /consultum

An open advisory communication channel for delegated AI workers — when a worker from other provider than Claude gets stuck, it escalates a curated problem to a stronger model of choice for a read-only second opinion.

Pher217 /LibreControl

LibreControl — one universal /remote-control for every coding agent you run (Claude Code, Codex, Gemini, Cursor, Copilot, any MCP tool), across all your machines, centralized into the single chat app you choose: Telegram, WhatsApp, Slack, Signal, Discord. Self-hosted, with approvals and audit.

rodrigorpassos /claude-codex-failover

Cross-provider failover between Claude Code (Opus) and Codex CLI (GPT): when one provider goes down, the other resumes the interrupted session.

ribato22 /volante

Volante — a transparent, user-owned model router and orchestration control plane (alpha). Plans a task DAG, filters hard capabilities, then routes each sub-task across your configured models (Anthropic, OpenAI-compatible, Ollama, Claude Code, Codex) with a full, auditable decision trace. No LangChain/CrewAI/LiteLLM.

keepwhatworks /trinity

Ask all three. Keep what works. A local-first cross-provider council layer for Claude, Codex, and Gemini — one prompt, all three answer, a chairman synthesizes the verdict. No new app, no API key; your transcripts never leave your machine.

Verhex /xerify

Verify before you trust. Cross-provider verification with LLMs and Jev. CLI, library & MCP.

anywave /lattice-consult-mcp

Cross-provider LLM consultation MCP server (OpenAI + Anthropic + Gemini + xAI Grok + Jan local). First concrete impl of Trellis DSM-26 AngularTechnique substrate.

hraness /ghostget

Retired

nauka-lm /llm-tool-recommendation-replication

Replication Package: Hallucination Mitigation in LLM-Based Tool Recommendation — A Cross-Provider Architectural Ablation Study Across Two Model Generations